The big picture: Closed model labs will benefit from closer scrutiny of open models.
- Critics, including Trump administration adviser David Sacks, say the scrutiny could amount to regulatory capture: Rules intended to improve AI safety could instead entrench the largest companies by making it harder for competitors to release models.
- Anthropic CEO Dario Amodei has argued that it is harder to keep open-weight models secure because once their weights are released, developers lose the ability to revoke access, update security guardrails, or prevent abuse.
- This is exactly what supporters of open-weight models see as their strength: no single company can control who uses them and how.
OpenAI views open source as part of the “democratization of access.” said a spokesman, albeit on a national scale. The company hasn’t outlined how that would work.
- OpenAI, which has its own open source model, has warned against distillation – a process in which a model is trained on another company’s model. This is a reference to senior Trump administration officials.
- U.S. Trade Representative Jamieson Greer said the agency views Chinese distillation as a form of intellectual property theft.
- Finance Minister Scott Bessent said yesterday that the same standards should apply to models from China as US models and would be tested in the coming weeks.
Friction point: There are no detailed safety standards for U.S. models yet, and the government is in the process of figuring out what they should look like.
- If U.S. officials’ claims about distillation are confirmed, it would mean that Chinese model suppliers used American models to develop comparable systems at lower costs and then released subsidized versions in Western markets to compete with U.S. labs.
Between the lines: Advocates warn that treating distillation as IP theft could lead to de facto restrictions on open-weight models, including models developed independently.
- “Advances in Chinese open-weight models are not an argument against openness,” said the OpenAI spokesperson.
- “They reinforce the need for a coherent national framework that allows the US to quickly evaluate new models, manage risks and put the most powerful AI tools in the hands of cyber defenders,” the spokesperson added.
Zoom out: Proponents of open models say they are critical for companies looking to diversify their AI use. They are also crucial for scientific discoveries in this phase of AI expansion.
- “As a researcher, I am very concerned if, for example, we are denied access to all Chinese models,” Suresh Venkatasubramanian, a former Biden White House technology adviser, told Axios, adding that this would be a “big problem for research.”
- He argues that open models are particularly valuable to researchers because they can be modified and studied more freely than closed models.
Enlarge: OpenAI and Anthropic may have similar rules for AI regulation at the federal level, but their differences are more noticeable at the state level.
- OpenAI endorsed a Massachusetts border security bill on Tuesday that doesn’t go as far as the version of the bill that Anthropic endorsed weeks earlier.
- The Anthropic-backed version required independent model testing for disaster risks and public reporting of the results every six months. This version also authorized the attorney general to sue companies for noncompliance and impose civil penalties.
- The bill advocated by OpenAI does not provide for this.
What we see: OpenAI CEO Sam Altman will meet with lawmakers and the White House next week to discuss upcoming releases of new models.
https://www.axios.com/2026/07/22/openai-anthropic-open-models-trump-china
